Antimicrobial resistance (AMR), the emergence of bacterial strains that survive antibiotic treatments, is a major global public-health concern, causing >1.2 million deaths yearly. If AMR is left unchecked, this figure will exceed 10 million by 2050.It is thus crucial for clinicians to know whether an illness is (i) caused by a bacterial infection, (ii) what species of bacteria is causing the infection, and (iii) what antibiotics would be effective in treatment. This information will optimise the prescription of the most appropriate antibiotic and minimise inappropriate antibiotic use. Improving the turnaround times of current tests will facilitate better outcomes for individual patients and support antimicrobial stewardship initiatives, thus reducing selection pressures for AMR at the individual and population-level.
Our work on AMR detection combines state-of-the-art microfluidics with single-cell microscopy and image analysis (based on machine learning) to rapidly detect and identify bacterial pathogens in a range of clinical specimens, and to determine to what degree they are susceptible to a range of antibiotics. Our approach uses fluorescent sensors to identify the species and resistance profile of single cells within a clinical sample. We have developed a microfluidics platform that captures bacteria from low-density samples with near 100% efficiency and can be used in combination with MER-FISH to identify bacterial species and mixed infections. Once captured, the resistance phenotypes of these bacteria be classified by deep learning models. This method reaches >80% accuracy in classifying single cells for a range of antibiotics, which results in very high accuracy in determining the resistance profile of an entire sample.
This project is highly collaborative and utilises long-standing relationships with the Modernising Medical Microbiology group at the John Radcliffe Hospital (1, 2). To enable translation of our assays into a clinical diagnostics platform, we are exploring a variety of research questions related to species identification, antibiotic response, and device development.
In 2023, our AMR team launched a citizen science project called Infection Inspection on The Zooniverse. This project attracted more than 5,000 volunteers, who learned how to identify antibiotic-resistant E. coli and contributed more than 1 million classifications. These data are helping us to understand why some antibiotic-treated cells are more likely to be misclassified than others. This project may also be re-opened in the future, to ask additional research questions(3).
